One SQL across lakes & warehouses

Companies are pitching a simpler way to analyze scattered data: one SQL query that reaches into Snowflake, Databricks and Amazon Web Services files without first building a separate pipeline for each system. (youtube.com) The demo at the center of this claim was posted in early April 2026 by Coffing Data Warehousing, which said its Nexus software joined Snowflake, Databricks and Amazon Web Services files in JavaScript Object Notation, Apache Parquet, comma-separated values and Excel formats in “one query.” (youtube.com) Structured Query Language is the standard language analysts use to ask databases questions, while Extract, Transform, Load tools usually copy and reshape data before those questions can run. Snowflake’s own documentation says data integration tools are commonly used for preparation, migration and warehouse automation. (docs.snowflake.com) A data warehouse like Snowflake stores data in a managed analytics system, while a lakehouse like Databricks runs analytics closer to files in cloud storage. Databricks says its platform connects to cloud object storage and external data systems, including through Lakehouse Federation and catalog federation. (docs.snowflake.com) (docs.databricks.com) Amazon Simple Storage Service, or Amazon S3, is the file layer in this setup. Amazon Web Services says analytics tools can query JavaScript Object Notation and column-based Apache Parquet files in S3, and Parquet is designed for faster retrieval than row-based formats like JavaScript Object Notation. (docs.aws.amazon.com 1) (docs.aws.amazon.com 2) The pitch behind tools like Nexus is federation: leave data where it already lives and send the query across systems instead of copying everything into one place first. Databricks says query federation is suited to on-demand reporting and proof-of-concept Extract, Transform, Load work, while catalog federation can support longer hybrid setups. (learn.microsoft.com) That approach is not unique to one vendor. Databricks documents federated querying into Snowflake, and Snowflake documents native connectors for Databricks Spark workloads, including automatic query pushdown that decides which processing runs on which system. (learn.microsoft.com) (docs.snowflake.com) The tradeoff is that “one query” does not mean one engine or one security model underneath. Databricks says federated access still requires network connectivity, credentials, catalog objects and permissions such as CREATE CONNECTION and CREATE FOREIGN CATALOG in Unity Catalog. (docs.databricks.com) (learn.microsoft.com) Snowflake makes a similar distinction between querying data already loaded into Snowflake tables and querying files that remain in external cloud storage through external tables and stages. In Snowflake’s documentation, the source of truth for external-table data remains in the cloud storage system, not in Snowflake itself. (docs.snowflake.com) So the real shift is not that SQL suddenly learned new tricks in April 2026. It is that vendors and third-party tools are increasingly packaging cross-platform access as a point-and-click workflow, with the promise that analysts can join warehouse tables and cloud files without standing up a fresh Extract, Transform, Load stack first. (youtube.com) (docs.databricks.com)